Chat with Ling 3.0 Flash VL.

Ant Group’s inclusionAI lab’s Ling 3.0 Flash with sight added. Only 5.5B of its 124B works on each word, so it is very cheap, and it reads pictures and video.

Start chatting with Ling 3.0 Flash VL now — free, no account

At a glance

  • Maker: inclusionAI
  • Released: 2026
  • Parameters: 124B
  • Active per token: 5.5B
  • Context window: 131k
  • Size: Large
  • Memory: Handles long documents
  • Good at: Very cheap picture and video reading
  • Supports: tools, images, zero data retention
  • License: Apache 2.0
  • Upstream API price: $0.18 per million tokens

What is Ling 3.0 Flash VL?

Ling 3.0 Flash VL is the cheapest model on UsingOpen, and one of the most interesting: Ant Group's inclusionAI lab built the Ling line around a mission — making capable AI cheap enough for everyone — and Flash VL is that mission at its purest. The VL stands for vision-language: the model reads pictures and video as well as text.

The economics come from aggressive sparsity: 124 billion parameters in total, of which only about 5.5 billion wake up per token — one of the leanest active ratios on the site. The trade is the context window: 131k tokens, roughly 100,000 words, modest by the standards of this catalogue. At $0.18 per million tokens, tied with DeepSeek's V4 Flash as the cheapest option here, it is also the only bargain-tier model with eyes.

It supports tools, is covered by zero-data-retention terms, and its stated speciality is exactly what the price suggests: very cheap picture and video reading. If your workload is look at this and tell me what it is, at volume — screenshots, product photos, video frames — this is the model the catalogue was built around.

What can you do with Ling 3.0 Flash VL?

Cheapest model on the site, with eyes — the use cases write themselves:

  • Bulk image questions: what is in this picture, at thousands of pictures per dollar.
  • Video frame analysis — extracting what happened, clip by clip.
  • OCR-adjacent work: reading screenshots, receipts, signs and scanned pages.
  • Cheap vision pipelines where a pricier multimodal model would blow the budget.
  • Everyday text questions too, at $0.18 per million tokens.
  • Tool workflows on a strict budget, thanks to tool support at this price.

How does Ling 3.0 Flash VL compare?

Against DeepSeek's V4 Flash — its co-holder of the $0.18 floor — it is sight versus scale: the DeepSeek is text-only but carries a million-token window, while Ling trades context for picture and video reading. Between them they cover the bargain basement from opposite ends.

Against Xiaomi's MiMo V2.6 Flash ($0.28), the nearest thing in spirit, MiMo adds audio input and a million-token window for ten cents more per million tokens. Ling's answer is the same as inclusionAI's mission statement: cheapest capable model, widest access.

Is Ling 3.0 Flash VL right for you?

Ling 3.0 Flash VL is the pick when the job is looking at things, cheaply, at volume. At $0.18 per million tokens with picture and video input, it is the bargain-basement sighted model on the site, and for pipelines that ask what is in this image a few thousand times a day, nothing else comes close on cost. The 5.5B-active design is what makes the price possible, and the vision-language speciality is what makes it useful rather than merely cheap.

Respect the trade-offs. The 131k context window means long archives go elsewhere; text-heavy work is better served by DeepSeek V4 Flash at the same price with a million-token reach; and audio belongs to the MiMo models. Ling is a specialist — a vision specialist at a price that makes specialisation affordable — and it is happiest when used as one.

It is also the cheapest way to experiment with multimodal pipelines before committing real money: prototype with Ling, and if the concept works, promote the pipeline to Gemma 4 26B or MiMo Flash for polish. inclusionAI's whole thesis is that capable AI should be cheap enough to try; this model is that thesis with a camera attached.

Frequently asked questions

What does the VL in Ling 3.0 Flash VL mean?

Vision-language. The model reads pictures and video as well as text — the sighted version of Ant Group's Ling 3.0 Flash design. Vision-language models read pixels and text together, which is what lets it answer where the total is on a receipt, not merely what the receipt says.

What is the context window of Ling 3.0 Flash VL?

131k tokens — roughly 100,000 words. Modest by this catalogue's standards: fine for documents and conversations, not the place to dump a whole archive the way the million-token models allow.

Why is Ling 3.0 Flash VL so cheap?

Aggressive sparsity — only about 5.5 billion of its 124 billion parameters are awake per token — plus a lab with a mission. inclusionAI is Ant Group's effort to make capable AI cheap enough for everyone, and $0.18 per million tokens is the mission in numbers. The lab’s name is the mission statement — inclusionAI exists to make capable models affordable, and Ling is the line where that shows up in the invoice.

Who makes Ling 3.0 Flash VL?

inclusionAI, the open-model lab of Ant Group, the Chinese fintech giant. The Ling family is the lab's open-weight line, released under Apache 2.0. Ant Group’s scale gives the lab an unusual design assumption: models meant to be cheap at volume, because volume is the world they are built for.

Does Ling 3.0 Flash VL support image input?

Yes — pictures and video. That is the VL in the name, and at $0.18 per million tokens it is the cheapest sighted model on UsingOpen.

Is Ling 3.0 Flash VL good for text-only work?

It handles text fine, but if your work is purely textual, DeepSeek V4 Flash matches the $0.18 price with a million-token window instead of Ling's 131k. Ling's edge is specifically cheap vision. The split at the $0.18 floor is clean — Ling for sight, DeepSeek for scale — and between them the bargain tier covers both shapes of cheap work.

Is Ling 3.0 Flash VL free to chat with on UsingOpen?

Yes. Pick it in the model picker and start chatting — no account, no payment, with history saved only on your device. It is also the cheapest model on the site to simply try: a full session of vision questions costs roughly nothing.

Related models on UsingOpen

Also worth a look if you are weighing Ling 3.0 Flash VL up:

  • MiMo V2.6 Flash by Xiaomi: the quick Xiaomi model — 15B of 309B awake, reading pictures, video and sound.
  • DeepSeek V4 Flash by DeepSeek: the bargain coder — 13B of 284B awake, a million-token window at $0.18 per million tokens.
  • Gemma 4 26B by Google: the mixture-of-experts Gemma — 25B total, 3.8B awake, quick on a budget.

Head-to-head comparisons

See how Ling 3.0 Flash VL lines up with the models people most often weigh it against:

Model availability changes as upstream capacity comes and goes — the picker on the chat page shows what is ready right now.

Chat with Ling 3.0 Flash VL free on UsingOpen — no account, no sign-up

← All open-weight models · Compare models · Free AI chat · FAQ